| Definition | Framework for testing RL algorithms in a multi-agent exploration scenario, in unknown areas. |
| No of environments | 2 |
It provides a framework for testing RL algorithms in a multi-agent exploration scenario, in unknown areas. Each agent has a limited field of vision by an emulated LiDAR sensor, and agents may or may not have a limited communication range. There is a limit of 4 agents at the moment, that can exchange map information when in communication range. It has an implementation of Dueling Double Deep Q Network (DDDQN), where one agent is training the model, while the rest are dummy agents. In evaluation episodes, the learnt model is loaded into all agents.